Robin Szemeti <[EMAIL PROTECTED]> writes:
> On Wed, 28 Mar 2001, you wrote:
> >
> > On Wed, Mar 28, 2001 at 09:26:38PM +0100, Robin Szemeti wrote:
> >
> > > (my pseudo-transaction scheme for MySQL is basically : .. do this and
> > > return a closure to undo it if I to .. bung the closures in an array ..
> > > if something screws up then back it all off by walking along the array
> > > and executing the closures ... its not rocket science but it works ..
> > > sort of .. I used it for doing multiple inserts into a spread of tables
> >
> > I did something similar. It worked too, until not only did an insert
> > fail, but when I was backing out, a delete failed too. There was much
> > head-scratching. A week later, the hard disk died and the head-scratching
> > stopped.
>
> ;)) ..
>
> > Unfortunately, if you implement this sort of thing, mysql loses it's only
> > advantage over other databases - speed. But I wasn't allowed to upgrade
> > to (eg) postgresql for silly reasons which I forget now.
>
> well .. since in most web based uses of MySQL the 99% of queries are
> simple 'select * from blah where something=something_else' .. the
> speed is all you need .. every now and again there is reason to add
> a user or, very occasionally, someone buys something .. and those
> bits have the pseudo-transactions in .. yeah .. its slow, but I'd
> ratehr have that bit slow and the rest lightning quick than pretyy
> much anything else ..
But the *REALLY IMPORTANT* uses of the database are the ones where
you're moving money about and doing order fulfillment. And guess what,
those *must* be transactional.
--
Piers
